RESUMO
OBJECTIVES: We developed a new algorithm to identify high-risk patients for underweight after oesophagectomy for cancer. Patients were assigned to an age-gender-specific body mass index percentile (AG-BMI) which is then used in a survival analysis. This model was able to identify patients more at risk for being underweight in comparison with the classically used BMI. It shows a worse overall survival (OS) in patients with a preoperative AG-BMI < 10th percentile. The aim of this study is to validate this new model based on a cohort of patients from an external high-volume institution specialized in oesophageal cancer surgery. METHODS: The validation cohort consists of 407 patients operated on between 1999 and 2012 with the prerequisite data to calculate AG-BMI and OS. The base cohort consisted of 642 consecutive patients, operated on in our institution between 2005 and 2010. Age, gender, height and weight on the day before surgery were used to calculate the BMI and the AG-BMI. OS was analysed and a multivariate analysis was performed. RESULTS: Incidence rates of the AG-BMI < 10th percentile risk-patients in the validation cohort showed similar results to our original results (17.8 vs 17.2% for the base cohort) with a similar significant OS difference between at-risk patients and not-at-risk patients (P < 0.001). Multivariate analysis found the same five independent prognosticators for OS in both datasets: age, early versus advanced disease, resection status, number of positive lymph nodes and the AG-BMI 10th percentile, but not BMI itself. In the validation cohort, gender was identified as an additional independent prognosticator. The worse OS survival in AG-BMI < 10th percentile in both patient populations was related to a significantly higher number of deaths without oesophageal cancer recurrence. CONCLUSIONS: This study validates the newly developed AG-BMI model to predict more accurately a subgroup of patients at risk for worse survival after oesophagectomy. Improved perioperative identification of risk factors for poorer OS could help to develop perioperative strategies to reduce these risks.